Does artificiality matter? Comparing working memory performance for artificial and real faces

Artificial faces are frequently used in research on perception, attention, working memory, and long-term memory. Although artificial faces offer advantages in terms of standardization, time, and cost, people tend to recognize real human faces more accurately than artificial faces in long-term memory tasks. However, there is no direct comparison of working memory performance between human and artificial faces. This study aims to examine the effect of artificial faces on working memory performance. Additionally, the study investigated the effects of the emotional valence of facial stimuli and the cognitive load of the task on working memory performance.

Forty-five adults (18-36 years) completed an n-back task with two conditions: human faces and computer-generated versions of the same identities (artificial faces). Working-memory load was manipulated at two levels (1-back and 2-back), and emotional valence was varied across three levels (neutral, happy, and angry). Sensitivity (d') and reaction times served as dependent measures to examine the effects of stimulus type, cognitive load, and emotional valence on working-memory performance.

Results showed that sensitivity was lower and reaction times were longer for artificial faces compared to the real faces. Moreover, performance decreased as cognitive load increased, whereas the effect of emotional valence was not statistically significant.

These findings suggest that artificial faces are processed less efficiently not only in long-term memory but also in working-memory. The results are discussed within the context of the face-space model and the other-race effect.